[6291] in s-news-athena
Reject unimodality??
daemon@ATHENA.MIT.EDU (Wenzl, Thurman)
Fri Jan 27 17:43:14 1995
From: "Wenzl, Thurman" <TYW1@NIOSHE2.EM.CDC.GOV>
To: 'snews' <s-news@utstat.toronto.edu>
Date: Fri, 27 Jan 95 17:14:00 EST
Did folks see this brief exchange (excerpted below) on stat-l. I'm looking
for an implementation of something like this in S+; any leads?
I know how to use ms() - barely - to estimate the parameters of a
bimodal-appearing distrib - but I'm looking for something 'quick and dirty'
to reject unimodality, prior to going on with ms() etc.
Thanks in advance,
Thurman Wenzl, NIOSH Cinti, tyw1@nioshe2.em.cdc.gov
Multimodality is, indeed, an intuitively simple concept which proves tricky
to
quantify systematically. You might check out the following two references:
Larkin, R. P. (1979). An algorithm for assessing bimodality vs. unimodality
in
a univariate distribution. Behavior Research Methods & Instrumentation,
11(4),
467-468.
Hartigan, P. M. (1985). Computation of the Dip statistic to test for
unimodality. Applied Statistics, 34(3), 320-325.
The latter reference includes FORTRAN code.
--------
Andy Gilpin, Dept. of Psychology, |Internet: andy.gilpin@uni.edu
Univ. of Northern Iowa, Cedar Falls, |Phone: (319) 273-6104
Iowa 50614-0505 |
